Learning to Coordinate with Experts

Mohamad H. Danesh, Tu Trinh, Benjamin Plaut, Nguyen X. Khanh · 2025

Read Learning to Coordinate with Experts by Mohamad H. Danesh, Tu Trinh, Benjamin Plaut, Nguyen X. Khanh (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-09583.txt.

Description

Mohamad Danesh and co-authors build a test bed for AI agents that must know when to hand a task to an expert. Their 2025 paper calls the test bed YRC-Bench. In real use, agents will meet tasks past their skill, and blind push can cause harm. Handing off helps, but expert help costs time and money, so timing the call matters. The hard case they stress gives no chance to learn from the expert during training.

The test bed spans many settings with a shared interface, fake experts, and a scoring path. It ships with known baseline methods. The authors use a split of proposer and checker roles to probe how methods fare and to guide next work.

Catalog record

CREATORMohamad H. Danesh, Tu Trinh, Benjamin Plaut, Nguyen X. Khanh
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE68 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-09583

Browse the full library · About the Tower of Babel